A Construction Method of Target Detection Adaptive Model Based on Cyclegan and Pseudo-label

Abstract

The invention discloses a method for constructing an adaptive model for target detection based on CycleGAN and pseudo-labels, comprising: S1, preprocessing of source domain data set and target domain data set; S2, using CycleGAN network to convert the source domain data set into an approximate The intermediate domain dataset of the target domain dataset, and input the intermediate domain dataset into the Faster R‑CNN network for training to obtain a preliminary domain adaptive model Q; re-input the target domain dataset into the model Q to obtain pseudo-labeled Target domain data set; S3, the intermediate domain data set and the target domain data set with pseudo-labels are alternately input into the model Q for iterative updating and optimization, and finally the target detection domain adaptive model based on CycleGAN and pseudo-labels is obtained. The present invention uses the confidence to improve the target detection total loss function of the Faster R-CNN network to train the target detection domain adaptive model, which can solve the problem of domain drift in target detection due to distribution differences between the two domains.

Description

technical field [0001] The invention relates to the technical field of deep learning, in particular to a method for constructing an adaptive model for target detection based on CycleGAN and pseudo-labels. Background technique [0002] Existing target detection methods based on deep neural networks (such as AlexNet, VGGNet, GoogleNet, and ResNet, etc.) can make the learned model apply to the test set when the data distribution of the training set and the test set is strictly consistent, or it can be used as a test set. Obtain higher detection accuracy. However, when the model trained by the training set is deployed in the actual natural scene, due to the fact that the actual natural scene environment is often uncontrollable, such as the huge difference in object appearance, background, lighting, climate, image quality, etc., the difference between the two is There is a difference in the data distribution, which leads to a significant drop in the detection accuracy of the mod...

AI Technical Summary

Problems solved by technology

[0003] The technical problem to be solved by the present invention is to provide a method for building an adaptive model for target detection based on CycleGAN and pseudo-labels for the problem of domain drift in target detection due to distribution differences between the two domains. Improve the target detection total loss function of the Faster R-CNN network to train a target detection domain adaptive model based on CycleGAN and pseudo-labels

Images